[jonahgabriel]

Cause

When production extraction withholds invalid JSON but grades raw provider text, HandlerQualityGateIntent returned only the evidence mismatch (DELIVERABLE_EVIDENCE_MISMATCH: cleaned content does not match the declared extraction evidence) and withheld the invalid label (INSTRUMENTEMENT_DEFECT).

Change

node_delegation_quality_gate_reducer, HandlerQualityGateIntent and handler_quality_gate._evaluate_response_contract preserve the existing mismatch failure and append the existing extractor contract_failure_reasons, so the invalid label now accompanies the mismatch. No new node, script or wire field.

Tests

New focused cases invalid, valid, invalid_then_valid and valid_mismatched_evidence run through typed ModelQualityGateIntent; existing terminal tests still prove refusal count and provider enum. RED first, same focused file: 2 failed, 15 passed, with the missing schema/token diagnostic for invalid and invalid_then_valid; the valid positive control passed. GREEN: 18 passed in 3.90s. Ruff check and format check passed for both changed files; git diff --check passed.

Limitation

The focused handler replay uses a captured token, not a new real deployed delegate failing/passing an enum run. No runtime deployment; no new wire fields. Commit, push, PR creation, repository hooks, ledger and landing handoff belong to the launching host under remote instructions.

Lab: host=h202 lane=lab-fill-omn18989-10081430 command=env -u PYTHONPATH uv run pytest tests/unit/delegation/test_declared_label_set_membership_omn18989.py -q observed=18 passed

File-scoped pre-commit: ruff and ruff-format passed. Full repository hooks and CI will run on the launching host.

Evidence-Ticket: OMN-18989

[github-actions]

✅ Architectural Review — PASSED

Errors: 0
Warnings: 0

What this checks

Rule	Description
ARCH-TOPIC-001	No hardcoded Kafka topic strings in handler code
ARCH-DI-001	No event_bus=None bypass
ARCH-DI-002	No direct Handler instantiation outside workflow_runner/adapters
ARCH-DI-003	No reinvented DI containers
ARCH-TOPIC-002	contract.yaml topics follow `onex.{cmd

---

No architectural violations found.

Static architectural lint — no model inference (OMN-14176).

[github-actions]

✅ Hostile Reviewer — PASSED

Blocking findings (critical/error): 0
Total findings: 7
Models succeeded: qwen3-review,gpt-oss-review

---

Gate semantics

Verdict	Meaning	Blocks merge?
passed	>=2 models succeeded, no critical/error findings	No
blocked	CRITICAL or ERROR findings found	Yes
degraded	Fewer than 2 models succeeded (infra unavailable/timeout)	Yes (OMN-15110)

Powered by omniintelligence.review_pairing.cli_review — node-based adversarial review via HandlerLlmCliSubprocess (OMN-8524)

[jonahgabriel]

Lab: head=7d50f0b7df20c8c45cb9bcc5f1658bf3798ae1c9 host=h200 lane=omn17427-3f28 exit=0
Command: env -u PYTHONPATH uv run pytest tests/unit/delegation/test_declared_label_set_membership_omn18989.py tests/unit/delegation/test_quality_gate_response_contract_omn15193.py tests/unit/delegation/test_reasoning_preamble_veto_omn18379.py tests/ci/test_repo_evidence_caller_workflow.py -q -p no:cacheprovider
Observed: 70 passed in 6.28s.
Head 7d50f0b adds two commits: a merge of origin/dev that keeps dev's version 0.4.306 in pyproject.toml and uv.lock (the branch's 0.4.305 bump conflicted with dev), and a change so _evaluate_response_contract calls extract_deliverable once and the DELIVERABLE_EVIDENCE_MISMATCH return reuses extraction.contract_failure_reasons.
The second commit answers the hostile reviewer's blocking finding (redundant extraction call, agreed by qwen3-review and gpt-oss-review on the previous head).
The Tests (Split 2/20) red on the previous head was tests/ci/test_repo_evidence_caller_workflow.py failing on contracts/OMN-20728.yaml (AC2, AC3 unbound) from the base; dev bound them in 9a235b6, and the merge carries that.